Why climate models disagree
What this page covers
The structural reasons climate models give different answers, how that uncertainty splits across three distinct sources, and what the hot-model problem is.
The three sources of uncertainty
To be written. Hawkins & Sutton 2009 framing — model uncertainty, scenario uncertainty, internal variability. For short-lead projections (next 20 years) internal variability dominates; for long-lead (end of century) scenario uncertainty dominates; model uncertainty dominates the middle.
Model uncertainty — structural differences
To be written. Each model makes different choices about convection, clouds, ocean mixing, sea ice, vegetation. Knutti & Sedláček 2013; Knutti et al. 2013 (model genealogy).
Scenario uncertainty — the SSPs
To be written. SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5 — what each represents; how the choice cascades into projection magnitude.
Internal variability — the climate’s own noise
To be written. The climate system has natural variability (ENSO, NAO, etc.). Single-model large ensembles surface this; CMIP6 r1i1p1f1 doesn’t.
The “hot model” problem
To be written. Sources: Hausfather et al. 2022 Nature; Sherwood et al. 2020; Forster et al. 2021 (AR6 Ch 7). CanESM5 ECS = 5.62°C, outside AR6 very-likely range. Why this matters for impact studies. The constrained-projection approach (AR6) vs the binary exclusion approach (Atlas AFR-13). Tebaldi 2022 counter-argument.
What this means when you cite the Atlas
To be written. Practical guidance: cite the ensemble mean as central estimate; cite the ensemble SD as uncertainty; choose AFR-13 default for headline numbers, FULL-18 for sensitivity / upper-tail risk framing.